Year: 2026
Type: other
Abstract: Engaging citizens in municipal decision-making and local governmental processes is important. There are various platforms for information shared by the municipalities in Finland, but they are too scattered, complicated, and overwhelming for the average citizen. AI has become increasingly more common in the past few years, and could be utilized to help citizens become more involved in their municipality. This thesis aims to test how Large language models (LLMs) could be used to gather the scattered information from various platforms in the Finnish municipality of Malax into one place and make it more accessible. While at the same time, allow for personalization, translation, and summarization to make the information more understandable and suited for each individual. The thesis presents a prototype of a personalized feed with an integrated chatbot. The thesis covers the methods for scraping the data and structuring it into a Neo4j knowledge graph database, a personalized feed built using Streamlit, and a chatbot built in Python using a multi-agent GraphRAG system for retrieving and answering user questions based on the data in the knowledge graph. Furthermore, the thesis compares performance between different LLM models and providers, translation accuracy, and factual accuracy of the chatbot. The personalized feed and chatbot are also given to ten Malax citizens to obtain feedback and opinions on the prototype. The findings of this thesis show that the prototype implementation functions well for gathering scattered, complicated municipal information and making it more accessible, which enables citizens to be more involved in their local government. However, there are still improvements that could be made to further enhance the user experience and LLM consistency. The prototype also does not take into account hosting, security, and scalability, which would be required for future iterations.
Source: Doria (University of Helsinki)
Author: Ena, David
Language: English
Field: Social Sciences
Domain: Social Sciences
Open Access status: green
OpenAlex ID: W7165237415